Pulmonary embolism (PE) is a common, costly, and potentially lethal disease in the US. While patients with PE are almost universally treated as inpatients and their length of hospital stay (LOS) varies widely, there is evidence that outpatient care with low-molecular-weight heparins or early discharge of hospitalized patients are effective and safe options for up to 50% of patients with PE. However, physicians may be reluctant to treat patients with PE on an outpatient basis or with brief hospitalization when the perceived risk of mortality or experiencing a short-term adverse outcome is not well quantified. Therefore, the broad objective of this project is to develop a clinical prediction rule that accurately identifies low-risk patients with PE who are potential candidates for outpatient care or early discharge. The specific aims are 1) to derive a clinical prediction rule that identifies patients with PE who have a very low 30-day mortality risk (<= 2%), and 2) to assess the performance of this rule in predicting other relevant outcomes (i.e., major bleeding, respiratory failure [RF], cardiogenic shock, cardiac arrest) or processes of care (i.e., thrombolysis, mechanical ventilation [MV], ICU admission, LOS, hospital readmission). This is a retrospective cohort study using 3 large and reliable databases, the PHC4 database, the MediQual Atlas database, and the National Death Index (NDI). These databases contain a rich set of clinical information and outcomes data. Our prediction rule for prognosis will be derived from 8,000 patients discharged with a diagnosis of PE from all acute care hospitals in Pennsylvania during the calendar years 2000-2001. Patients with PE will be identified using primary ICD-9-CM discharge codes for PE or secondary discharge codes for PE coupled with primary discharge codes that represent complications or treatments of PE. Baseline data that include 35 potential clinical predictors of short-term mortality in PE will be abstracted from the PHC4/Atlas databases. The primary outcome will be 30-day mortality ascertained from the NDI; secondary outcomes will be major bleeding, RF, cardiogenic shock, cardiac arrest, thrombolysis, MV, ICU admission, LOS, and readmission. Classification tree analysis will be used to construct a simple clinical prediction rule that identifies a 20% subgroup of all patients with a 30-day mortality rate of 2% or less. The predictive accuracy of this rule will be externally validated in an independent cohort of 4,000 patients with PE from calendar year 2002 using identical patient identification strategies as for the derivation cohort. The safety of this rule will be tested in the validation cohort by computing the proportion of very low-risk patients who die within 30-days or have another adverse outcome. This innovative application will derive a clinical prediction rule for prognosis that has the potential to improve the cost-effectiveness of the management for PE. The long-term goal of this project will be to validate and implement this rule in prospective studies to test its safety and effectiveness and to establish future admission/early hospital discharge recommendations for patients with PE.